1. Field of the Invention
The present invention relates to a color space quantization descriptor structure for performing quantization of a color space in order to search a multimedia based on contents when multimedia feature information is color information, in particular to a color space quantization descriptor structure comprising quantization description information which has a tree structure divided into a plurality steps of superior color spaces and subordinate color spaces, and the tree structure recursively includes a plurality of subordinate color spaces.
2. Descriptor of the Prior Art
In the conventional technology, in order to search a multimedia based on contents, particularly in order to search a still image or a moving image, the search is performed by using feature information of the image such as color information, texture information, shape information or time-change information of the image.
In order to search the image, using the color information for describing the characteristics of an image is very useful for searching the multimedia based on the contents.
In particular, describing the color information of one pixel or group of pixels is the basic method for constructing a digital image.
When the color information is described on the each pixel (or group of the pixels), a color element value is differed in accordance with a certain color space (for example, HSV (Hue Saturation Value) color space, RGB (Red Green Blue) color space etc.) to represent the characteristics of an image in more effective way according to its usage.
As described above, in order to use the color information of the image for searching the multimedia based on the contents, a process on the color information of the each pixel according to a certain method is required. As a representative method there is a color histogram for grasping distribution of the color information displayed on the overall images.
In order to construct the color histogram, dividing the color space into a plurality of parts is required, that is the quantization of the color space.
The color space quantization is to divide the color space which is represented as a three dimension or a four dimension etc. into subordinate color spaces, it uses such characteristics that colors which are adjacent to each other within the color space looks similar to eyes of human.
The color space quantization is needed for searching the still image or moving image based on the color information, and the color space quantization is useful in speed efficiency and storage efficiency of the search.
In the speed efficiency aspect, when the all color information which is allotted to the each pixel are used in matching of the image, it is difficult to perform the matching and it increases computing complexity.
Accordingly, the whole image or the part of it is described by the color histogram. Herein, when the color histogram is formed by the information of the quantization of color space, the amount of color information used for the matching is smaller than the original image where the color information is allotted to the each pixel, accordingly the computing complexity is decreased.
In addition, in the storage efficiency aspect, the storage space allotted to the each image is reduced by the compaction of the information.
There is a uniform quantization, a non-uniform quantization, and a lookup table for performing the quantization in the conventional method.
The uniform quantization method divides the color space into the subordinate color spaces with a uniform interval on each axis constructing the color space.
Process of the uniform quantization method is very simple, and it is efficient in the speed aspect when the two images represented as the uniform quantization are matched. On the contrary, when the original color space has non-uniform characteristics, the characteristics of the color space is not reflected well, accordingly the search performance becomes worse.
Meanwhile, the non-uniform quantization divides the color space into the subordinate color spaces with a non-uniform interval on the each axis constructing the color space. The non-uniform quantization method is inferior to the uniform quantization method in terms of complexity, but it is superior in terms of searching performance because it performs the non-uniform quantization of the color space which shows non-uniform characteristics in itself.
In addition, the lookup table quantization method directly describes the upper limit and lower limit of the axis of each subordinate color space, which doesn't need any quantization operation required for the uniform quantization and non-uniform quantization.
The lookup table quantization method is very efficient in the search ability or speed aspect, but a search apparatus should always keep the lookup table requiring large amount of storage.
Accordingly, the color space quantization method which can reflect well the substantial characteristics of the color space, is efficient in terms of complexity, and have high search performance is required.
In addition, the color space quantization descriptor structure which can describe the color space quantization information precisely and simply and, at the same time, is compatible with various search engine is required. However, as described above, the conventional color space quantization technology performs the quantization with the quantization information of the final one step obtained by simply dividing the color space uniformly or non-uniformly, accordingly the substantial characteristics of the color space can not be reflected properly.
In addition, the conventional color space quantization technology performs the quantization with the quantization information of the final one step obtained by simply dividing the color space uniformly or non-uniformly, accordingly it is not efficient in terms of the quantization speed or matching speed.
In addition, the conventional color space quantization technology performs the quantization with the quantization information of the final one step obtained by dividing the color space uniformly or non-uniformly, accordingly it can not have high search performance in the multimedia search based on content.
In addition, the conventional color space quantization technology performs the quantization with the quantization information of the final one step obtained by dividing the color space uniformly or non-uniformly, accordingly it can not simply and accurately describe the color space quantization information while making it compatible in the various search engine.